Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence.

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Tue, 28 Apr 2026 04:46:06 GMT

quantensalat@scicomm.xyz — Tue, 28 Apr 2026 04:46:06 GMT

@wronglang @devsimsek Yes, sure. I mean I can imagine it improving somewhat still, like when you augment your training set for image recognition by adding noise to a smaller set, but only to a point before it goes downhill from feedback.

No, my gut feeling is rather that there have to be much more effective ways to train a model than to brute force funnel billions of pages of text to a transformer which blindly fits relations between words and structures without understanding them, that seems like doing it the hard way, even if I'm not expert enough to tell you what an alternative would look like

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Tue, 28 Apr 2026 04:38:34 GMT

troed@masto.sangberg.se — Tue, 28 Apr 2026 04:38:34 GMT

@resuna

Everything in your post was wrong - so why did you post it?

@devsimsek

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Tue, 28 Apr 2026 04:38:15 GMT

urban_hermit@mstdn.social — Tue, 28 Apr 2026 04:38:15 GMT

@devsimsek
"Touch grass." It is not just a reminder to take a break or get some fresh air.

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Tue, 28 Apr 2026 04:35:45 GMT

quantensalat@scicomm.xyz — Tue, 28 Apr 2026 04:35:45 GMT

@wronglang @musicman @devsimsek No, agreed, more compute with the same type of model and the same training data sounds totally unplausible to me as a long term strategy

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Tue, 28 Apr 2026 04:19:28 GMT

calcifer@masto.hackers.town — Tue, 28 Apr 2026 04:19:28 GMT

@devsimsek you have an awkward sentence here you might want to know about: “Even though I like to say yes, i neither have the enough research nor I want to comment on it”

I think you’re going for something like “even though I’d like to say yes, I have neither enough research nor any desire to comment on it”… but I’m not entirely sure.

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Tue, 28 Apr 2026 04:11:12 GMT

brahmabelarusian@regenerate.social — Tue, 28 Apr 2026 04:11:12 GMT

@devsimsek This & overall the bigger issue of forced overinclusion & attempted hyperteliance on machine learning systems, mostly done by governments & their private partners, like autoshutoff on cars, chatbots as talk therapists& biometric ID/digital ID instead of regular ID card systems, is destined to fail.... It's not so much that activists will win in court or public protests on how these things at least mostly violate civil liberties & are based on data & intellectual property theft.... It's that fundamentally none of these systems actually work!

They couldn't even write a specific mechanism or method for the vehicle one because nothing fitting the mandate has been developed & the nearest ones obviously dont work.

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Tue, 28 Apr 2026 03:07:03 GMT

onekind@beige.party — Tue, 28 Apr 2026 03:07:03 GMT

@devsimsek I'd be interested to see the same analysis of human consciousness. It is well understood that complexity is a regime on the absolute edge of chaos.

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Tue, 28 Apr 2026 02:15:00 GMT

rootwyrm@weird.autos — Tue, 28 Apr 2026 02:15:00 GMT

@anne_twain @devsimsek a better equivalence explanation.

Here is a 'smart hammer.' It promises to never smash your thumb. And between 20 and 60% of the time, it works! The other 80 to 40% of the time it explodes and takes off your entire arm and sets the nearest three houses on fire.

The question is not "why are people not stopping when it explodes" or "how do we filter the explosions."
The question is "WHY THE FUCK ARE PEOPLE STILL USING AN EXPLODING HAMMER?!"

I need to remember this one.

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Tue, 28 Apr 2026 02:08:34 GMT

rootwyrm@weird.autos — Tue, 28 Apr 2026 02:08:34 GMT

@anne_twain @devsimsek there is no process. There is no intelligence. There never was and there never will be.
It's a bad stochastic parrot written by children who should have been flunked out of 7th grade math and 3rd grade English as illiterate. Used and pushed by people who aren't capable of reviewing a fast food order, or even placing one.

And guess what? All irrelevant because it takes an incomprehensible level of stupidity to even use a tool that fails dangerously constantly.

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Tue, 28 Apr 2026 00:51:14 GMT

mike805@noc.social — Tue, 28 Apr 2026 00:51:14 GMT

@musicman @Quantensalat @devsimsek Anyone who ever copied an audio tape (or worse a VHS tape) knows that the copy is always worse than the original. And in the video case, soon unwatchable.

Ever heard a repeating echo on a video meeting that just turns to a buzz? Same phenomenon.

So what you need is an AI that can perform experiments in the real world to learn how to do better whatever it is you want it to do.

Inbreeding animals doesn't work too well either.

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Mon, 27 Apr 2026 23:26:55 GMT

emma@orbital.horse — Mon, 27 Apr 2026 23:26:55 GMT

@devsimsek so it doesn't get stuck in a local optimum, it hill-climbs a non-existent one?

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Mon, 27 Apr 2026 23:10:42 GMT

wronglang@bayes.club — Mon, 27 Apr 2026 23:10:42 GMT

@Quantensalat @musicman @devsimsek depends on what you mean by far fetched, certainly nothing as easy as "their more compute at it' which is what made this jump in investment so dramatic.

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Mon, 27 Apr 2026 23:08:18 GMT

wronglang@bayes.club — Mon, 27 Apr 2026 23:08:18 GMT

@Quantensalat @devsimsek the main issue is that unless you maintain an external signal (so human input in the form of token sequences that are actually carefully curated for coherence) the models become more and more incoherent. Sounds like you're on board with that. The next step is that we're quickly devaluing money spent on human creativity and the world is awash in LLM garbage. So the human signal *is* disappearing.

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Mon, 27 Apr 2026 22:59:45 GMT

resuna@ohai.social — Mon, 27 Apr 2026 22:59:45 GMT

@rootwyrm @devsimsek

Mark V. Shaney.

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Mon, 27 Apr 2026 22:59:19 GMT

resuna@ohai.social — Mon, 27 Apr 2026 22:59:19 GMT

@troed @devsimsek

Large language models are fundamentally different from mammals on every level. They do not build models or reason about them. A rat is more "intelligent".

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Mon, 27 Apr 2026 22:57:51 GMT

resuna@ohai.social — Mon, 27 Apr 2026 22:57:51 GMT

@rootwyrm @dpiponi @Quantensalat @devsimsek

"How can a system with no reference make correct determinations? Simple: it can't."

Especially since it has no model of "correctness" other than "similar to the symbol streams the neural net weights were initialized from".

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Mon, 27 Apr 2026 22:45:28 GMT

anyia@lgbtqia.space — Mon, 27 Apr 2026 22:45:28 GMT

@devsimsek "Don't worry bro, we can totally fix this by adding a committee of expert LLMs to reason about what training data to select, another committee of LLMs to plan the optimal training order, and then a larger one to evaluate the training output. We just need you to sign this cheque for our next three hyperscale GPU data centres..."

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Mon, 27 Apr 2026 21:28:13 GMT

aburka@hachyderm.io — Mon, 27 Apr 2026 21:28:13 GMT

@devsimsek did an LLM write this toot or do LLMs just write like you

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Mon, 27 Apr 2026 21:18:01 GMT

troed@masto.sangberg.se — Mon, 27 Apr 2026 21:18:01 GMT

@devsimsek The existence of humans disprove the paper.

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Mon, 27 Apr 2026 20:44:36 GMT

rednikki@toot.boston — Mon, 27 Apr 2026 20:44:36 GMT

@devsimsek “slowly forgets what reality looks like.” Sort of like billionaires.

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Mon, 27 Apr 2026 20:16:38 GMT

dpiponi@mathstodon.xyz — Mon, 27 Apr 2026 20:16:38 GMT

@Quantensalat @devsimsek For something more formal on this subject see

https://arxiv.org/abs/2601.03220

The abstract starts "Can we learn more from data than existed in the generating process itself?"

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Mon, 27 Apr 2026 20:11:28 GMT

lorxus@yiff.life — Mon, 27 Apr 2026 20:11:28 GMT

@devsimsek @qualia I think you claim too much here. As I understand it, this result deals only with the intrinsic failures of RL-flavored approaches and not things like self-play, let alone problems that might arise from merely very good AI that still outdoes humans economically.

And I largely agree! I'm glad that someone's finally formalized the intuition that synthetic data is sawdust to bulk out real-world data with and more carefully investigated catastrophic forgetting and the general weaknesses of gradient descent.

That said... to what extent did you have Claude write this post? Because the format is... distinctive.

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Mon, 27 Apr 2026 19:43:25 GMT

hermlon@yuustan.space — Mon, 27 Apr 2026 19:43:25 GMT

@devsimsek isn't the idea of self-improving AI that the AI modifies its code, so the underlying algorithm / architecture?

Reply to Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence. on Mon, 27 Apr 2026 19:33:51 GMT

paul@notnull.space — Mon, 27 Apr 2026 19:33:51 GMT

@devsimsek excellent. Thanks for the overview!